دانشگاه آزاد اسلامی واحد تهران جنوب

صفحه اصلی / نخستین همایش هوش مصنوعی و پژوهش های نوظهور: همگرایی انسان و سیستم های هوشمند

Verifiable and Privacy-Preserving Large Language Model Interactions: Addressing Blockchain-based Accountability Frameworks

نویسندگان :

Navid Homayouni¹ Razieh Farazkish²

1- دانشجوی دکتری تخصصی ، و مدعو دانشگاه آزاد اسلامی واحد تهران جنوب 2- دانشگاه آزاد اسلامی واحد تهران جنوب

کلمات کلیدی :

LLM،Blockchain،ZKPs،authenticity

چکیده :

Large Language Model (LLM) chatbots are becoming ubiquitous, yet mechanisms for ensuring the accountability and verifiability of their interactions remain nascent [1]. Blockchain technology offers potential solutions through its immutable and transparent ledger capabilities, enabling the creation of trustworthy interaction logs [2]. However, storing sensitive chat data directly on a blockchain fundamentally conflicts with user privacy requirements [3]. This study emerging landscape of frameworks that integrate blockchain with Privacy-Enhancing Technologies (PETs) to address this challenge, aiming to provide verifiable accountability without compromising confidentiality [4]. We analyze architectures based on hashing, encryption, and zero-knowledge proofs (ZKPs), comparing their trade-offs regarding verifiability, privacy guarantees, scalability, and computational overhead. Key challenges, including blockchain throughput limitations, PET computational costs (especially ZKPs), key management complexity, and the inherent privacy-verifiability tension, are discussed [5]. We identify significant research gaps and propose future directions focusing on Layer-2 scaling solutions, advancements in efficient ZKP systems, hybrid blockchain models, user-centric controls, and formal verification methods [6]. This work concludes that while significant hurdles exist, the synergy between blockchain and PETs presents a viable path towards building responsible and trustworthy LLM interaction systems